Interaction in Vivo and in Vitro of the Metastasis-inducing S100 Protein, S100A4 (p9Ka) with S100A1*

The calcium-binding protein S100A4 (p9Ka) has been shown to cause a metastatic phenotype in rodent mammary tumor cells and in transgenic mouse model systems. mRNA for S100A4 (p9Ka) is present at a generally higher level in breast carcinoma than in benign breast tumor specimens, and the presence of immunocytochemically detected S100A4 correlates strongly with a poor prognosis for breast cancer patients. Recombinant S100A4 (p9Ka) has been reported to interact in vitro with cytoskeletal components and to form oligomers, particularly homodimers in vitro . Using the yeast two-hybrid system, a strong interaction between S100A4 (p9Ka) and another S100 protein, S100A1, was detected. Site-directed mutagenesis of conserved amino acid residues involved in the dimerization of S100 proteins abolished the interactions. The interaction between S100A4 and S100A1 was also observed in vitro using affinity column chromatography and gel overlay techniques. Both S100A1 and S100A4 can occur in the same cultured mammary cells, suggesting that in cells containing both proteins, S100A1 might modulate the metastasis-induc-ing capability of S100A4. transformant was determined using a liquid b -galactosidase assay (24). The number of Miller units of b -galactosidase activity was calculated using the for- mula: Miller units 5 (1,000 3 optical density of the reaction)/(volume of cell assayed in ml 3 time in min 3 optical density of the culture at 600 nm). Mutagenesis of S100A4— Amino acid residues phenylalanine 72 and tyrosine 75 of the human S100A4 protein sequence (25) were both altered to glutamine by changing the triplets from TTC and TAC to

S100 proteins are a family of low molecular weight, acidic proteins that contain two distinct EF-hands with different affinities for calcium (1). Elevated levels of one S100 protein, S100A4, are closely associated with the process of metastasis in breast and other cancer cells in rodent animal models and in human cancer specimens. S100A4 or its mRNA is found at an elevated level in metastatic relative to non-metastatic rat (2) and mouse (3) tumor cell lines and benign relative to malignant human breast tumors (4). Elevation of the level of rat (5) or human (6) S100A4 in benign rat mammary tumor cells results in the acquisition of metastatic capability by some of the cells. In transgenic mouse models of breast cancer, elevated levels of S100A4 in neu oncogene-induced (7), or in murine mammary tumor virus-induced (8), benign mammary tumors yield lung metastases. In colorectal adenocarcinoma specimens, elevated levels of immunocytochemically detected S100A4 are associated with the more malignant carcinomatous regions of the primary tumors and with liver metastases (9). The precise interactions whereby S100A4 induces metastasis are not fully understood. S100A4, or its mRNA, is found in many normal cells, some of which are motile (10,11), and S100A4 has been associated immunofluorescently with the actin/myosin cytoskeleton in fixed cells (5,(12)(13)(14). Calcium-dependent interaction of S100A4 with actin (15), tropomyosin (16), and non-muscle myosin (17)(18)(19)(20) has been reported in vitro. S100 proteins form homodimers and heterodimers (21,22); S100A4 forms calcium-and dithiothreitol-independent homodimers in vitro (23). The conditions employed for these interactions in vitro cannot match the conditions inside the cell, where functional interactions are likely to take place. Thus, the yeast two-hybrid system has been utilized to find interactions in vivo for the metastasis-inducing protein S100A4.

EXPERIMENTAL PROCEDURES
DNA Constructs for the Yeast Two-hybrid System-For screening a human breast tumor two-hybrid library constructed with pYESTrp2 vector (Invitrogen, Groningen, Netherlands), a bait vector was constructed by excising the LexA cassette from the pHybLex/Zeo vector (Invitrogen), and inserting it into the pAD-GAL4-2.1 vector (Stratagene, La Jolla, CA) to produce the LexA vector with the LEU2 selectable marker. A 303-base pair cDNA corresponding to the coding region of human S100A4 (p9Ka) mRNA was obtained by PCR 1 of an expression vector pET-p9Ka 2 and was subcloned into the LexA vector to produce the LexA-S100A4 bait construct. Nucleotide sequencing ensured the integrity of the cloning and sequences. Selection for this bait plasmid in yeast cells was on leucine-free plates or medium.
Yeast Transformation and Lift Assays and Liquid ␤-Galactosidase Assays-For the screening of the breast cancer two-hybrid library, yeast strain L40 (genotype: MATa his3⌬200 trp1-901 leu2-3112 ade2 LYS2::(4lexAop-HIS3) URA3:: (8lexAop-lacZ)GAL4) cells were transformed with LexA-S100A4 to produce a stable L40-LexA-S100A4 strain. A culture was transformed with the human breast cancer pYESTrp2 library DNA according to the lithium yeast transformation protocol (Invitrogen). Yeast transformants were selected by their growth on leucine-, tryptophan-, and histidine-free plates and media. Lift assays (according to Stratagene's protocol) were carried out 7 days after transformation, and colonies that gave a blue color within 16 h were isolated.
For each individual clone, both the bait plasmid, LexA-S100A4, and the target plasmid from the YESTrp2 library were isolated from the cells. The recovered YESTrp2 target plasmids were cotransformed into L40 yeast cells along with the original bait plasmid, LexA-S100A4, and lift assays were carried out to reconfirm the interaction in the yeast cells.
The activity of ␤-galactosidase of each positive transformant was determined using a liquid ␤-galactosidase assay (24). The number of Miller units of ␤-galactosidase activity was calculated using the formula: Miller units ϭ (1,000 ϫ optical density of the reaction)/(volume of cell assayed in ml ϫ time in min ϫ optical density of the culture at 600 nm).
Mutagenesis of S100A4 -Amino acid residues phenylalanine 72 and tyrosine 75 of the human S100A4 protein sequence (25) were both altered to glutamine by changing the triplets from TTC and TAC to CAG using the PCR extension method of site-directed mutagenesis (26). The two external primers used were: 5Ј-XhoI primer, TCTCTCGAGCT-GCTGTCATGGCGTGC and 3Ј-PstI primer, TGCGCCTGCAGTG-GAGTTTTCATTTCTTC. The inner primers for the F72Q mutation were: 5Ј-GACCAGCAAGAGTACTGTGTCTTCC and 3Ј-CTCTTGCTG-GTCCACCTCGTTGTCCC. The inner primers for the Y75Q mutation were: 5Ј-GAGCAGTGTGTCTTCCTGTCCTGC and 3Ј-GACACACT-GCTCTTGGAAGTCCACC. The cDNAs with the point mutations were digested with XhoI and PstI and inserted into the LexA vector. DNA sequencing confirmed the nucleotide changes and the coding frames for fusion protein expression.
Recombinant Proteins-Rat recombinant S100A4 (rat rS100A4) was produced and purified as described previously (12). A recombinant human S100A1 cDNA was obtained using PCR from a clone selected in the two-hybrid screen. The cDNA open reading frame, which encoded the 93 amino acids of human S100A1 protein, was subcloned into the expression vector pGEX-2T, and rS100A1 was produced as a fusion protein with 26 kDa of glutathione S-transferase (GST) protein at the NH 2 terminus (GST-S100A1). The fusion protein was purified to near homogeneity using glutathione-Sepharose 4B (Amersham Pharmacia Biotech). Histidine-tagged-rS100A1 (His-S100A1) was expressed in pET16b(ϩ) vector and purified with histidine-binding resin (Novagen).
In Vitro Binding Assay for S100 Proteins-For the detection of binding of S100A1 and S100A4 proteins in vitro, the GST-S100A1 fusion protein or GST alone was attached to glutathione S-Sepharose (Amersham Pharmacia Biotech) slurries at a concentration of 0.5 mg/ml of slurry. 100 l of a solution of 0.1 mg/ml rS100A4 protein in KTT buffer (140 mM KCl, 20 mM Tris-HCl, pH 7.4, 0.1% (w/v) Triton X-100, 5 mM dithiothreitol, with 1 mM CaCl 2 , or 1 mM EGTA) was incubated with 100-l aliquots of the GST or GST-S100A1 slurries for 3 h at 4°C. The mixtures were applied to spin columns and centrifuged in a microcentrifuge at 800 rpm for 1 min. The columns were spun-washed with the KTT buffer at least 10 times. Unbound proteins were eluted with at least 50 bed volumes of KTT until no further protein was eluted. Bound proteins were eluted with 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM dithiothreitol, 10 mM reduced glutathione, and 0.1% (w/v) Triton X-100. Eluted S100A4 was analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) (27) followed by transfer onto PVDF filters (Immobilon, Millipore, France), and incubation with a 1:500 dilution of an affinitypurified rabbit polyclonal antibody to rat rS100A4, produced as described previously (28). Bound antibodies were detected using the ECL system (Amersham Pharmacia Biotech).
For gel overlay, the recombinant GST-S100A1, His-S100A1, GST, and S100A4 proteins were subjected to SDS-PAGE (12.5% (w/v)) and electroblotted onto PVDF membranes. The membranes were incubated with overlay buffer (0.5% (w/v) bovine serum albumin, 0.25% (w/v) gelatin, 0.5% (w/v) detergent Nonidet P-40, 100 mM NaCl, 50 mM Tris, pH 7.5, containing 3 g/ml rat rS100A4 plus either 0.5 mM calcium ions or 1 mM EGTA) for 4 h and washed four times for 5 min each in the same buffer without the bovine serum albumin, gelatin, or S100A4. Membranes were incubated with blocking buffer (3% (w/v) bovine serum albumin, 0.1% (w/v) Tween 20, 2% (w/v) Marvel in Tris-buffered saline). The membranes were washed three times for 5 min each in wash buffer (0.1% (w/v) Tween 20, 2% (w/v) Marvel in Tris-buffered saline). The membranes were incubated with 2.5 ml of the wash buffer containing a 500-fold dilution of the rabbit anti-rat S100A4 serum for 1.5 h. After being washed in buffer six times for 5 min each, the filters were incubated in a 2,500-fold dilution of a secondary antibody for 1 h. After a final wash with Tris-buffered saline for 30 min, bound antibody was detected using the ECL system with exposure against Kodak XAR-5 film.
For RT-PCR, 1 g of total RNA or 0.5 g of poly(A)-containing RNA was reverse transcribed using Superscript according to the supplier's protocol (Life Technologies, Inc., Paisley, Scotland), and the resulting cDNA was amplified by 35 cycles of PCR using Taq polymerase (Life Technologies, Inc.) with the following human S100A1-specific primers: 5Ј-primer, ACAGGTCTCCACACACAGCTCC; 3Ј-primer, AGGCTC-GAGAGGAAGGGCGCTGC. PCR products were analyzed by electrophoresis on 1.4% (w/v) agarose gels and stained with ethidium bromide. Control amplifications omitted the reverse transcriptase.
Immunofluorescent Detection of S100 Proteins-About 500 cells were plated per well of chambered slides and cultured for 24 h. The cells were fixed with 4% (w/v) paraformadehyde in phosphate-buffered saline for 10 min and permeabilized with 0.1% (w/v) Triton X-100 in phosphatebuffered saline for 5 min. After a 30-min incubation in blocking buffer (2% (w/v) bovine serum albumin in phosphate-buffered saline), the cells were incubated with primary antibody, either rabbit anti-S100A4 (1:50 dilution) or mouse anti-S100A1 (1:20 dilution) in blocking buffer. After being washed four times with phosphate-buffered saline, the cells were incubated with secondary antibody for 1 h. Fluorescein isothiocyanateconjugated anti-mouse IgG for anti-S100A1 staining and tetramethylrhodamine ␤-isothiocyanate-conjugated anti-rabbit IgG for S100A4 staining were used. After washing, the slides were dried and mounted. Photographs were taken using a Zeiss LSM 510 confocal microscope.

RESULTS
Interaction of S100A4 in Yeast Cells-To identify heterologous interactions in vivo between S100A4 and components of malignant/metastatic cells, more than 2 ϫ 10 6 yeast cell transformants from a two-hybrid cDNA library, constructed from mRNA isolated from a human breast cancer specimen, were transfected with the S100A4 (p9Ka) cDNA bait vector. Target plasmids were recovered from 14 independent blue colonies, and purified plasmid DNAs were cotransformed individually into L40 yeast cells along with the S100A4 (p9Ka) bait plasmid. Reisolation of plasmid DNA and cotransformation were repeated serially three times, after which only four colonies showed the blue color in lift assay on each occasion. Plasmid DNA was isolated from these four transformants, and the nucleotide sequences of the cDNAs in the target vectors were determined. The sequence of each of the four target plasmids was precisely that of human S100A1. The four colonies containing both the S100A4 bait and S100A1 prey plasmids exhibited ␤-galactosidase activities that were 20 -70-fold higher than colonies containing only bait plasmid or only a prey plasmid (Fig. 1a). However, the prey constructs in the library (Invitrogen) contained an additional 19 amino acids of protein sequence inserted between the B42 activating domain and the NH 2 terminus of the target protein. To rule out the possibility that this extra sequence might be interacting directly with the S100A4 protein, the extra sequence was deleted from one of the isolated target clones. L40 strain yeast transformants containing both the pYESTrp2ExS100A1 prey construct, bearing the deletion, and the S100A4 bait construct LexA-S100A4, yielded a blue color within 3 h in the lift assay, and the ␤-galactosidase activity of the deleted clone was greater than that of the original clone containing the additional 19 amino acids insert (Fig.  1b). Thus, the ␤-galactosidase activity was not caused by the presence of these additional amino acids at the NH 2 terminus of the cloned S100A1 protein in the yeast cells from the breast cancer library. The above experiments suggest that, at least in yeast cells, S100A4 can interact in a heterologous manner with S100A1.
Point Mutation in the S100A4 cDNA Interrupts the Interaction between S100A4 and S100A1 in Vivo-Two single nucleotide point mutations of the S100A4 cDNA were generated which yielded amino acid substitutions in residues involved in the dimer interface reported previously for S100A6 (34). These changes, which resulted in conversion of phenylalanine 72 and tyrosine 75 of the S100A4 protein sequence to glutamine, individually fully (Phe-72), or partially (Tyr-75), disrupted S100A4 dimer formation in the yeast two-hybrid system (not shown). Bait plasmids containing cDNAs bearing either of these mutations completely failed to yield blue colonies when cotransfected with S100A1 prey plasmids under the same conditions that yielded blue colonies in 6 h when unmutated S100A4 was used. These specific mutations individually abolished the interaction between S100A4 and S100A1 in the yeast cells as determined by ␤-galactosidase activity (Fig. 1c).
Interaction between S100A4 and S100A1 in Vitro-Antibodies directed against recombinant S100A4 or S100A1 were tested for their specificity toward their target recombinant proteins (Fig. 2). Using Western blotting, anti-S100A4 did not cross-react against NH 2 -terminal polyhistidine-tagged S100A1, recombinant GST, nor against a GST-S100A1 fusion recombinant protein, but did recognize S100A4 protein at the same loading. Anti-S100A1 recognized the His-tagged S100A1 and the GST-S100A1 fusion protein but did not cross-react onto S100A4 protein at the same loading (Fig. 2).
To find out whether it is possible to detect interaction between S100A4 and S100A1 in vitro, the GST-S100A1 fusion protein and the control GST were immobilized on Sepharoseglutathione columns. The interaction in vitro with recombinant S100A4 was detected by the retention of the S100A4 on the Sepharose-glutathione-GST-S100A1 column. After extensive washing, the bound proteins were eluted from the columns with excess glutathione, and the proteins in the resulting fractions were analyzed by SDS-PAGE and Western blot with anti-S100A4 antibody. When a sample of recombinant S100A4 was passed through a column containing bound recombinant GST, no reduced glutathione-elutable S100A4 was retained on the column beyond the wash fraction (Fig. 3c). However, when the S100A4 was passed through a similar column but containing immobilized GST-S100A1 fusion protein, rS100A4 was retained on the column, even after extensive washing with buffer, and was eluted, along with the GST-S100A1 fusion protein, with reduced glutathione. This result suggests strongly that S100A4 can bind to GST-S100A1 fusion protein but not to the GST control in vitro. Exactly the same result was obtained when the experiment was carried out with a cell lysate of Escherichia coli BL-21 cells which had been induced to produce S100A4 with IPTG (Fig. 3b). The interaction in vitro of purified recombinant GST-S100A1 with S100A4 (purified or nonpurified) was detected in the presence of calcium ions and could be reversed in the presence of 1.0 mM EGTA (Fig. 3b).

S100A4 (p9Ka) Interactions in Vivo
columns above were utilized, with buffers containing between 0 and 1.0 mM calcium ions or 0.5 or 1.0 mM EGTA. Interaction between S100A1 and S100A4 could be detected using the buffer with no added calcium; however, the amount of S100A4 re-tained on the GST-S100A1 column increased as the concentration of calcium in the buffer was increased (Fig. 4). 0.5 mM EGTA reduced the amount of S100A4 retained, and no detectable S100A4 was retained with 1.0 mM EGTA. These results suggest that the interaction of S100A4 with S100A1 is not markedly calcium ion-dependent over the normal range of calcium concentrations. However, removing all of the calcium with a high concentration of EGTA created conditions that reduced the retained S100A4 to an undetectable level.
Interaction between S100A1 and S100A4 was also demonstrated using a gel overlay technique with purified rS100A1 and rS100A4 proteins (Fig. 5). Recombinant S100A4 protein binding to both His-S100A1 and GST-S100A1, but not to GST protein, was detected by the antibody to S100A4. S100A1 and S100A4 Are Present in the Same Cells-To find out whether S100A1 and S100A4 coexist in the same cells naturally, the presence of S100A1 mRNA was sought in clonal breast tumor cell lines of known S100A4 status using a PCRbased assay. A correctly sized ethidium bromide-stained band of DNA on agarose gels was obtained following RT-PCR using human S100A1-specific primers on RNA from human mammary cell lines, MCF-7, MDA-MB-231, and SK-Br-3. The latter two cell lines have been shown previously to contain high levels of S100A4 mRNA (4) (Fig. 6). Sequencing of the PCR products confirmed that they corresponded to the human S100A1 mRNA. No bands were obtained with any of the RNAs when RT-PCRs were carried out in the absence of the Superscript reverse transcriptase. Thus, the mRNAs for both S100A1 and S100A4 were present in these clonal cell lines. The presence of both S100A1 and S100A4 protein in the same cells has been shown by Western blotting (Fig. 7) and by dual labeling immunofluorescence on MDA-MB-231 cells (Fig. 8). In the immu- FIG. 3. Interaction between S100A4 and S100A1 in vitro. Recombinant GST (panels a and b, lanes 1, 2, 5, and 6; panel c, lanes 1-4) or GST-S100A1 recombinant fusion protein (panels a and b, lanes 3, 4, 7,  and 8; panel c, lanes 5, 7, 8, and 9) were purified and immobilized on glutathione-Sepharose columns. Either a bacterial extract containing recombinant S100A4 (panels a and b; shown in lane 10) or purified S100A4 (panel c; shown in lane 10) was passed through the columns using a Tris buffer (panels a and b) or a phosphate-buffered solution (panel c), as described under "Experimental Procedures." In panels a and b, the buffer contained either calcium (lanes 1, 2, 7, and 8) or EGTA (lanes 3-6). The proteins that passed through the columns (designated PT) were collected (panel c , lanes 1 and 9). The columns were washed until no more protein was eluted, and the washes were either pooled (panel c, lanes 2 and 8, designated MW for mixed wash), or a final wash fraction was passed through (panels a and b, lanes 1, 3, 5, and 7, designated W). Proteins remaining bound to the columns were eluted, designated by E (panels a and b, lanes 2, 4, 6, and 8; panel c, lanes 3 and  7), as described under "Experimental Procedures." Samples of the fractions were subjected to SDS-PAGE as described under "Experimental Procedures," and a gel was stained (panel a) or the proteins were transferred onto membranes (panels b and c). S100A4 was detected by a rabbit anti-S100A4, which did not cross-react with S100A1 (panels b and c).  lane 4), the GST-S100A1 recombinant fusion protein (lane 5), and recombinant S100A4 (lane 10, designated A4) were also subjected to electrophoresis and Western blotting. m, monomeric; n, multimeric; n/a, not applicable. The multiple weak bands in panel a, lanes 4 and 8, arise from degradation of the GST-S100A1 fusion protein, and the apparent doublet for S100A4m in panel b is an overloading artifact.

FIG. 4. Effect of calcium ions on the interaction in vitro
between S100A4 and S100A1. GST-S100A1 fusion recombinant protein was purified and immobilized on glutathione-Sepharose beads. The beads were incubated at 4°C overnight with rS100A4 protein in Tris buffer containing no added CaCl 2 (lanes 1-3) or added CaCl 2 to 0.1 mM (lanes 4, 5, and 7), 0.5 mM (lanes 8 -10), 1.0 mM (lanes [11][12][13] or EGTA added to 0.5 mM (lanes 14 -16) or 1.0 mM (lanes 17-19). The mixtures were then applied to spun columns. After extensive washing, the GST-S100A1 was eluted from the beads with glutathione. The fractions, P (pass-through), W (last wash), and E (elute) were subjected to SDS-PAGE and Western blotting. The S100A4 that had been retained on the column was detected with anti-S100A4. The smear in the pass-through fractions arises from aggregation of the excess recombinant S100A4 protein. Molecular mass markers (M r ) of standard proteins on the stained gel are shown in lanes 6 and 20. nofluorescent experiments, both S100A1 and S100A4 were localized to the perinuclear region and to the cytoskeleton, and the fluorescence of each could be removed by preincubating each antibody separately with its cognate, but not the other recombinant protein (Fig. 8). Superimposition of the staining shows their partial colocalization on stress fibers and in the perinuclear region. DISCUSSION S100A4 is shown to interact with S100A1 in Saccharomyces cerevisiae cells and also in vitro. Although it has been reported previously that S100A4 forms homodimeric forms (23,35), this is the first report of S100A4 interacting with another member of the S100A family of EF-hand-containing proteins. Based upon the time taken for the blue dye arising from ␤-galactosidase activity to become visible, the interaction between S100A4 and S100A1 in vivo was more effective at stimulating reporter gene activity than the homodimeric interaction of S100A4 in the same system. S100A4 molecules that had been specifically mutated in conserved amino acid residues, reported previously to be associated with the dimer interface of S100A6 (calcyclin) (34), and which prevented S100A4 homodimerization in the yeast two-hybrid system, also failed to interact with S100A1, strongly suggesting that this latter interaction resembles the natural association of other S100 proteins into homo-(34) and hetero-(36) dimeric forms.
Homodimer and heterodimer formation of S100 proteins has been predicted from the three-dimensional structures of some S100 proteins and homology analysis (34). Natural homodimeric and heterodimeric association of S100A family members is well documented. Thus, S100A1 (37), S100B (22), S100A4 (23), and S100A6 (34) have been shown to form homodimers. Heterodimerization of S100A8 with S100A9 in neutrophils (21,36) and of S100A1 with S100B (38) in neural tissue has been described, and the formation of heterodimers between S100B and S100A6 (39) in human melanoma has been reported recently.
The heterodimerization of S100A1 and S100A4 reported here might be expected to have biological significance if these S100 proteins occur in the same cells. These proteins have been shown to coexist in the same cultured human breast cancer cells and in rat mammary cells, Rama 29 (not shown) (30), and coexpression of S100A1 and S100A4 has been reported previously in primary cultures and cell lines of human vascular smooth muscle cells (14). In the present experiments, S100A1 and S100A4 are both located in the perinuclear region and on the stress fibers of the cytoskeleton. Localization of S100A4 (5,12,14) and of S100A1 to the cytoskeleton (14) and of S100A4 to the sarcoplasmic reticulum (14) has been reported previously. FIG. 5. Interaction in vitro between S100A1 and S100A4 using gel overlay. Purified recombinant proteins, His-S100A1 (lane 1), GST (lane 3), rat S100A4 (lane 4), GST-S100A1 (lane 5), GST (lane 6), and human S100A4 (lane 7) were separated by SDS-PAGE (15% (w/v)) and blotted onto PVDF membranes. The membranes were incubated with S100A4, and any binding of rS100A4 was detected by anti-S100A4 (see "Experimental Procedures"). The banding pattern of molecular weight markers is shown on the left side of the image with molecular masses in kDa indicated. Lane 2 contained no protein.  3 and 4), SK-Br-3 ( lanes 5 and 6), and MDA-MB-231 cells (lanes 7 and 8) was subjected to RT-PCR (lanes 4, 6, and 8) or control amplification in which the reverse transcriptase was omitted (lanes 3, 5, and 7), using primers specific for human S100A1, as described under "Experimental Procedures." A PCR without template was also carried out (lane 2). The resulting PCR products were subjected to agarose gel electrophoresis along with molecular size markers (lane 1), and the gel was stained with ethidium bromide. The image has been reversed. The band of S100A1 amplification product is indicated by the arrow. FIG. 8. Immunofluorescent localization of S100A1 and S100A4 in human mammary cells in culture. MDA-MB-231 cells were grown in chambered slides, fixed, and incubated with a mouse monoclonal anti-S100A1 and fluorescein isothiocyanate-conjugated antimouse IgG (panel A) or with a rabbit polyclonal anti-S100A4 and tetramethylrhodamine ␤-isothiocyanate-conjugated anti-rabbit IgG (panel B). Superimposition of the images in A and B indicates partial colocalization of the two S100A proteins around the nucleus and on stress fibers. Preincubation of the anti-S100A1 IgG with rS100A1 abolished the immunofluorescence caused by S100A1 (panel D) but not that caused by S100A4 (not shown), and preincubation of the anti-S100A4 with rS100A4 abolished the immunofluorescence resulting from S100A4 (panel E) but not that resulting from S100A1 (not shown). Panels A-C and the phase-contrast image (panel F) are of the same field. Magnification ϫ 280; bar ϭ 50 m.
These observations suggest that S100A1 and S100A4 might also be interacting inside the cell at these same locations.
Because S100A1 and S100A4 are capable of dimerizing, it is possible that they have a common intracellular target. Both S100A1 and S100A4 individually have been shown to interact with a range of non-S100 protein targets. S100A1 has been shown to interact with brain-specific C and muscle-specific A isoforms of aldolase (40), MyoD (41), and a dodecylpeptide of the actin-capping, Cap Z (42), F-actin (14), the calcium-release channel of skeletal muscles (43), and twitchin kinase of Caenorhabditis elegans (44). S100A4, in contrast, has been reported to interact in vitro with a more limited range of intracellular target proteins, in particular cytoskeletal actin (15), tropomyosin (16), and non-muscle myosin heavy chains (17)(18)(19)(20). However, the interaction with actin might arise from a tendency of S100A4 to aggregate (14). The interaction of S100A4 with myosin occurs at a site on the non-muscle myosin heavy chains which has been reported to be phosphorylated by protein kinase C in vivo and in vitro (20). All of these interactions have been determined with single purified S100A4 protein, and it will be important to find out whether these interactions in vitro can be modulated by the presence of S100A1. It is possible that the interactions of S100A4 with myosin, which might be associated with the protein's metastasis-inducing capabilities, could be modulated by interaction with cytoskeletonassociated S100A1 (14). In this regard it might be interesting to note that S100A1 and S100A4 are colocalized in the high S100A4-expressing cultured Rama 29 cells (30), which do not express a metastatic phenotype.
The distribution of S100A1 and S100A4 proteins in normal cells is much more diverse. Although S100A1 was originally found in brain, where it interacts with S100B, it is also found in heart, slow twitch skeletal muscle, smooth muscle, and kidney (45). In contrast, S100A4 is not found in significant quantities in brain, but it is distributed widely and specifically elsewhere in the body, including kidneys, lungs, thymus, and muscle tissues, at least in the rat (28). The observation that S100A1 interacts with S100A4 as well as with S100B, but that S100B (46), but not S100A4 (28), is found widely in brain tissue, raises the possibility that S100A1 has different S100 partners depending upon the S100 proteins that are expressed in the particular cells concerned.